/ChronusArtifact

Primary LanguagePythonMIT LicenseMIT

ChronusArtifact

Artifact for ACM SoCC '21

https://dl.acm.org/doi/10.1145/3472883.3486978

This repository contains the artifact for the ACM SoCC '21 paper "Chronus: A Novel Deadline-aware Scheduler for Deep Learning Training Jobs". It includes following 2 parts:

  • survey: The detailed statstical information of user survey

  • code: Python Implementation of Chronus.

Trace We Use

Helios traces (SenseTime) download from HeliosData.

Philly traces (Microsoft) download from philly-traces.

Citation

If you use this code or survey in your research, please cite this project.

@inproceedings{10.1145/3472883.3486978,
  author = {Gao, Wei and Ye, Zhisheng and Sun, Peng and Wen, Yonggang and Zhang, Tianwei},
  title = {Chronus: A Novel Deadline-Aware Scheduler for Deep Learning Training Jobs},
  year = {2021},
  publisher = {Association for Computing Machinery},
  address = {New York, NY, USA},
  url = {https://doi.org/10.1145/3472883.3486978},
  doi = {10.1145/3472883.3486978},
  booktitle = {Proceedings of the ACM Symposium on Cloud Computing},
  pages = {609–623},
  numpages = {15},
  keywords = {Deadline-aware Scheduler, Deep Learning Training, Cluster Management System, GPU Datacenter},
  location = {Seattle, WA, USA},
  series = {SoCC '21}
}